General Purpose Technologies
GPTs are a somewhat controversial concept but are a potentially useful framework for understanding the impact of technology on the economy.
Technological innovation is an important driver of economic growth, particularly General Purpose Technologies (GPTs) due to their broad applicability. The development of GPTs and the magnitude and timing of their impact are difficult to predict, but understanding GPTs remains important as they are a potential source of economic cycles and can reshape society.
Definition
GPTs are technologies that alter large parts of the economy and society due to their broad applicability. GPTs often have limited use initially, but as the technology improves it spreads throughout the economy, bringing about generalized productivity gains. While products often come to mind first, the category also includes processes and organisational systems. Example criteria for classifying GPTs include:
A single, recognisable technology
The technology initially has significant scope for improvement
Pervasive (eventually used across many sectors)
Creates many spill over effects
The criteria used to identify GPTs are subjective though, leading to debate about what technologies actually qualify as GPTs. Many GPT lists include dozens of technologies but only IT, electricity and steam are universally recognized. The GPT theory is also somewhat controversial as it has little heuristic power.
Examples
Due to the subjectivity of the classification criteria and debate about which technologies meet the relevant criteria, there is significant variation in lists of GPTs. The list below is fairly comprehensive, although not exhaustive. When technologies are not universally accepted as GPTs, like clothing or lean production, it is generally due to debate about the technology’s impact or the breadth of use.
Table 1: General Purpose Technologies
(source: Created by author using data from Wikipedia)
ARK Invest have developed their own list of GPTs, which they believe currently present investment opportunities. This list is indicative of the difficulties in classifying GPTs. 3D printing is currently too niche and will likely never be used widely enough to be considered a GPT. Similarly, immunotherapy and reusable rockets will never be used broadly enough to be considered GPTs. Autonomous mobility and to a lesser extent adaptive robotics could be considered applications of artificial intelligence, rather than their own GPTs. In a similar manner, mobile connected devices, cloud computing and IoT could be considered applications of computing rather than their own GPTs.
Table 2: General Purpose Technologies Identified by ARK Invest
(source: Created by author using data from ARK Invest)
Effect on Society
Due to their broad applicability, GPTs have the potential to significantly alter societies through their impact on economic and social structures, including the organisation of work, the management of firms, skill requirements, the location of industries and the required supporting infrastructure. For example:
The development of language increased the ability of people to communicate, leading to more complex societies. Language, both written and oral, has allowed societies to scale beyond Dunbar’s number and contributed to the formation of cities and nations.
Clothing provided people with the ability to survive in a wider range of climates, leading to increased migration.
The domestication of plants and animals led to larger permanent settlements and the increase in productivity enabled specialisation in society.
Writing improved the transfer of knowledge through time and space, allowing knowledge to compound more effectively over time.
The factory system increased productivity, bringing many items into an affordable price range for the average person. This allowed spending to shift towards categories that were previously considered luxuries, making industries like entertainment and tourism feasible.
The internal combustion engine provided an inexpensive and readily available source of power, contributing to suburbanisation and globalisation. It also enabled industrial agriculture, which reduced the labour intensity of agriculture and freed people to move into the manufacturing and services sectors.
Electricity provided an easy to distribute source of power and provided a platform on which other technologies could be developed.
The internet reduced transaction costs, helping to enable new types of economic activities.
Artificial intelligence is increasingly providing the ability to automate non-routine tasks, which previously weren’t considered amenable to automation.
While GPTs are generally expected to increase productivity, the introduction of a new GPT does not always cause productivity to increase, particularly in the short-term. GPTs take time to improve and diffuse, and spill over benefits may take time to occur. The conversion of inputs to outputs is a complex process, dependent on many factors and it therefore should not be surprising that technological innovation is not easily observable in a simplistic measure like Total Factor Productivity (TFP).
Figure 1: Solow Growth Model Treatment of Technology
(source: Created by author)
It should also be noted that TFP is a residual in the Solow growth model that only measures productivity gains in excess of investment. It is therefore possible for technology to provide productivity gains that do not show up in TFP.
Figure 2: Granular Treatment of Technology in the Production Function
(source: Created by author using data from ITIF)
Technological Knowledge - Existing base of technology.
Facilitating Structure - Capital and labour along with the necessary supporting infrastructure.
Policy Structure - The structure through which public policy operates.
Policy - Public policies which influence technology.
Knowledge compounds over time in a path dependent process, meaning that the introduction of a new technology cannot have unique predetermined results. In addition, economic agents involved in the development of technology face uncertainty as well as risk. Agents operating under risk are expected to make the same choice, whereas agents operating under uncertainty may make different choices. This further complicates the expected outcome when a new technology is introduced.
The benefits of technological innovation should be apparent from the continued growth in GDP per capita over time. If no new GPTs were introduced, it is likely that the number of derivative technological developments would eventually diminish. GPTs help to sustain the growth process but do not necessarily accelerate it. The continued growth of output per labour hour is demonstrative of continued productivity growth, even if this does not always show up in TFP.
Figure 3: US Output per Labour Hour and TFP
(source: Created by author using data from The Federal Reserve)
Output per labour hour growth could also be the result of capital deepening. There are diminishing returns to adding capital though and it is unlikely this is a significant driver of growth in developed economies.
Figure 4: Diminishing Returns from Capital Investments
(source: Created by author)
The diffusion of innovations is also an important determinant of the productivity growth rate as the benefit of a technology can be spread over a larger time period or compressed into a smaller time period.
The social rate of return can exceed the private one due to both vertical and horizontal externalities.
Vertical externalities are the result of GPTs benefiting complementary assets. This generally increases demand for the GPT, which increases demand for inputs and in time likely lowers the cost of inputs.
Figure 5: Vertical Externalities
(source: Created by author)
Horizontal externalities are the result of GPTs benefiting a broad set of application sectors. Broad applicability provides a large market which in turn leads to more rapid improvement in the technology.
Figure 6: Positive Horizontal Externalities
(source: Created by author)
Horizontal externalities can also be negative when a legacy technology becomes redundant. Use of the technology is likely to decline, leading to lower investment in the technology and a stagnation of the technology.
Figure 7: Negative Horizontal Externalities
(source: Created by author)
The introduction of a new GPT will often be associated with an initial economic slowdown, before any productivity gains are registered. This may be due to:
Capital obsolescence due to technological uncertainties
An obsolescence of old technologies and skills
Time required for development of new infrastructure
Learning costs
Co-ordination problems
Necessity of complementary technologies
The magnitude of the disruption caused by technology can be seen in the shift in employment between sectors over time, a process that can happen relatively rapidly under the right conditions. While it may be easy for employees to change jobs in some cases, in others there may be a significant skills gap which is not easily resolved.
Figure 8: Employment by Sector
(source: Created by author)
Information and Communication Technologies (ICT) have likely caused structural unemployment in recent decades as employees have lost jobs in industries like print media and struggled to find equivalent jobs.
Figure 9: Labor Force Participation Rate 25-54 Year Olds
(source: Created by author using data from The Federal Reserve)
ARK Invest combined these concepts into a qualitative assessment of GPTs. Their broad recognition of technologies as GPTs, an optimistic assessment of the impact of those technologies and an apparently optimistic assessment of the adoption of these technologies leads to the rather absurd chart below. Instead of the predicted economic boom of unprecedented proportions, we remain in a period of weak economic growth with anemic productivity growth.
Figure 10: Potential Economic Impact of GPTs Identified by ARK Invest
(source: ARK Invest)
Conclusion
GPTs are an important driver of economic growth, although the theory has been questioned due to the subjectivity of identifying GPTs and the theories weak heuristic power. GPTs are generally expected to contribute to economic growth through TFP growth, but the relationship between technology and productivity is complex and economic activity may actually be negatively impacted to begin with as the economy adjusts to the new technology. It has been theorised that the pace of innovation has been accelerating and that we are potentially entering an economic golden age as the productivity gains from these technologies becomes apparent. Many recent technologies have relatively narrow applicability though and are still developing, limiting their current economic impact.